Hackers use Anthropic’s AI model Claude once again
Anthropic announced Thursday that Chinese state-backed hackers used the company’s AI model, Claude, to automate about 30 attacks against companies and governments during a campaign in September, according to a report from the The Wall Street Journal.
Anthropic said as much as 80 to 90 percent of the attack was automated with AI, a higher level than previous hacks. It happened “literally with the click of a button and then with minimal human interaction,” said Jacob Klein, Anthropic’s head of threat intelligence. Newspaper. He added: “The human was only involved in a few critical choke points, saying, ‘Yes, continue,’ ‘Don’t continue,’ ‘Thanks for that information,’ ‘Oh, that doesn’t look good, Claude, are you sure?’ »
AI-based hacking is becoming more common, as is the latest strategy of using AI to accomplish the various tasks necessary for an attack to be successful. Google has spotted Russian hackers using large language patterns to generate commands for their malware, according to a company report released on November 5.
For years, the U.S. government has warned that China was using AI to steal the data of American citizens and businesses, which China has denied. Anthropic said Newspaper that he is convinced that the hackers were sponsored by the Chinese government. During this campaign, the hackers stole the sensitive data of four victims, but as in previous hacks, Anthropic did not disclose the names of the targets, successful or unsuccessful. The company said the U.S. government was not an effective target.
